2015 Way Too Early Predictions

[Trotsky]

Ridiculously early first pass TBRW prediction for 2015, based on the currently announced early departures:

1. Union (33 pts)
2. Colgate (29)
3. Cornell (26)
4. Quinnipiac (25)

5. Yale (24)
6. Dartmouth (21)
7. St. Lawrence (21)
8. Clarkson (21)

9. RPI (20)
10. Harvard (17)
11. Brown (15)
12. Princeton (12)

Code Select Expand


RS PS Imp Prior WRet Ret Avg10 WInc Inc Nieu Pred
Brown 17 0 -2 15 0.639 9.585 15.700 0.361 5.668 15.253 11
Clarkson 24 0 -4 20 0.637 12.740 21.400 0.363 7.768 20.508 8
Colgate 29 0 0 29 0.960 27.840 22.000 0.040 0.880 28.720 2
Cornell 26 0 0 26 0.621 16.146 27.300 0.379 10.347 26.493 3
Dartmouth 16 1 4 21 0.979 20.559 22.400 0.021 0.470 21.029 6
Harvard 16 0 1 17 0.922 15.674 21.900 0.078 1.708 17.382 10
Princeton 8 0 -2 6 0.549 3.294 19.500 0.451 8.795 12.089 12
Quinnipiac 28 0 -3 25 0.637 15.925 23.667 0.363 8.591 24.516 4
RPI 21 -1 1 21 0.522 10.962 19.200 0.478 9.178 20.140 9
SLU 18 0 2 20 0.361 7.220 21.600 0.639 13.802 21.022 7
Union 37 0 0 37 0.667 24.679 25.600 0.333 8.525 33.204 1
Yale 24 0 1 25 0.646 16.150 23.000 0.354 8.142 24.292 5


If Ghost leaves, Union would drop to 32 pts.

[marty]

Ridiculously early first pass TBRW prediction for 2015, based on the currently announced early departures:

1. Union (33 pts)
2. Colgate (29)
3. Cornell (26)
4. Quinnipiac (25)

5. Yale (24)
6. Dartmouth (21)
7. St. Lawrence (21)
8. Clarkson (21)

9. RPI (20)
10. Harvard (17)
11. Brown (15)
12. Princeton (12)

Code Select Expand


RS PS Imp Prior WRet Ret Avg10 WInc Inc Nieu Pred
Brown 17 0 -2 15 0.639 9.585 15.700 0.361 5.668 15.253 11
Clarkson 24 0 -4 20 0.637 12.740 21.400 0.363 7.768 20.508 8
Colgate 29 0 0 29 0.960 27.840 22.000 0.040 0.880 28.720 2
Cornell 26 0 0 26 0.621 16.146 27.300 0.379 10.347 26.493 3
Dartmouth 16 1 4 21 0.979 20.559 22.400 0.021 0.470 21.029 6
Harvard 16 0 1 17 0.922 15.674 21.900 0.078 1.708 17.382 10
Princeton 8 0 -2 6 0.549 3.294 19.500 0.451 8.795 12.089 12
Quinnipiac 28 0 -3 25 0.637 15.925 23.667 0.363 8.591 24.516 4
RPI 21 -1 1 21 0.522 10.962 19.200 0.478 9.178 20.140 9
SLU 18 0 2 20 0.361 7.220 21.600 0.639 13.802 21.022 7
Union 37 0 0 37 0.667 24.679 25.600 0.333 8.525 33.204 1
Yale 24 0 1 25 0.646 16.150 23.000 0.354 8.142 24.292 5


If Ghost leaves, Union would drop to 32 pts.

Counterpoint?

[Trotsky]

Counterpoint?

One of me is going to look brilliant.

[Jim Hyla]

Counterpoint?

One of me is going to look brilliant.

So, is it the digital you, or the analog you? That is the question.

[marty]

Counterpoint?

One of me is going to look brilliant.

So, is it the digital you, or the analog you? That is the question.

I was thinking we had a right/left dilemma but since I don't understand where those damn numbers came from I'm at a loss...

[ursusminor]

Counterpoint?

One of me is going to look brilliant.

So, is it the digital you, or the analog you? That is the question.

I was thinking we had a right/left dilemma but since I don't understand where those damn numbers came from I'm at a loss...

Lets see if I remember this from past years. Not that Trotsky hasn't changed it many times in the past. RS is how many points the team had last year. PS is an adjustment due to upsets in the playoffs (and a reminder to me that RPI plays like crap when it counts). Imp is another adjustment for improvement in the second half of the season compared to the first. Both of these are zero sum adjustments. Prior is the sum of the three columns to the left.

WRet is a complicated calculation of how much of the teams strength is returning. It combines separate factors for forwards, d-men, and goalies. Ret = Prior*WRet, a measure of how many of the teams points are returning next season.

Up to this point things basically make sense to me. Now Trotsky wants to add in a factor for incoming players (but not including anything about them per se). Simultaneously this incorporates a measure of improvement in returning players. Avg10 is the average of how many points the team has had for the last 10 years, i.e., a measure of how good the team has been historically. WInc is the fraction of the team's points due to players who left, i.e. WInc=1-WRet. Inc=Avg10*WInc. So Trotsky is assuming that incoming players will play at a rate typical of how the team has done historically. That is as good as one could do without actually knowing anything about incoming players.

Nieu=Ret+Inc, a measure of how many points the team will have next season. Inc isn't zero sum. (I have forgotten if this is due to Vermont leaving and QU taking their place.) The sum of all the Nieu values is not 22*12=264. It's close but not exact. Pred is an ordering based on Nieu.

[Trotsky]

Lets see if I remember this from past years. Not that Trotsky hasn't changed it many times in the past.

Touchy subject.  Given infinite time and resources I would go back, retro-apply all of my criteria to all of the data sets, figure out which have been most accurate, and stick with those.  If you're curious, here is an index page for all the past calculations.

RS -- Pts in RS (normalize to 22 GP if different, the Elephant Walk Epicycle)

PS -- +1 / -1 for each upset win / loss in an ECAC PS series (no consy)

Imp -- Second half improvement in pts over first half; difference / 2 and rounded down.

Prior: RS + PS + Imp, representing the comparative strength of the team's returning components

WRet -- Weighted returning percentage.  It's not complicated, just labor intensive.  The percentage of scoring among the top six Fs that is returning weighted by 3 + the percentage of scoring among the top 4 D that is returning weighted by 2 + the percentage of all G minutes that is returning, all divided by 6.  This represents the percentage of team performance to be attributed to current components.

Ret -- WRet * Prior.  The total number of points contributed by the returning components.

WInc -- Weighted incoming percentage.  1 minus WRet.  The idea is that a team's strength is a sum of its current configuration (returning components) and its baseline program strength, with the current component weighted by how much of the team strength is returning.  If a team returned everybody their traditional strength wouldn't matter at all.  Incoming strength is estimated as the average performance over the prior ten years.  Note that this still takes into consideration atypical recent success (or failure)  since the prior ten years includes the prior two or three.

Avg10 -- Average pts last 10 seasons.  (In Qpc's case, that's an avg of 9 seasons)

Inc -- WInc * Avg 10.  The total number of points contributed by the incoming components.

Nieu -- Ret + Inc.  Total expected points.

Pred -- Rank order of Nieu.

As Ralph points out, Nieu is not exactly normalized to a 22 game season, but it's really close (this year's error is .0024).

I think this is the best balance between methodological defensibility and practicality of collection that I've had, but I'm open to suggestions.

[Scersk '97]

Plus ça change, plus c'est la même chose.

The only squads to exchange "pods" are predicted to be Dartmouth and RPI.

[ursusminor]

If Ghost leaves, Union would drop to 32 pts.

He signed today.

[Trotsky]

If Ghost leaves, Union would drop to 32 pts.

He signed today.

Updated calculations:

Code Select Expand


[b]RS PS Imp Prior WRet Ret Avg10 WInc Inc Nieu Pred[/b]
Brown 17 0 -2 15 0.639 9.585 15.700 0.361 5.668 15.253 11
Clarkson 24 0 -4 20 0.637 12.740 21.400 0.363 7.768 20.508 8
Colgate 29 0 0 29 0.960 27.840 22.000 0.040 0.880 28.720 2
Cornell 26 0 0 26 0.621 16.146 27.300 0.379 10.347 26.493 3
Dartmouth 16 1 4 21 0.979 20.559 22.400 0.021 0.470 21.029 6
Harvard 16 0 1 17 0.922 15.674 21.900 0.078 1.708 17.382 10
Princeton 8 0 -2 6 0.549 3.294 19.500 0.451 8.795 12.089 12
Quinnipiac 28 0 -3 25 0.637 15.925 23.667 0.363 8.591 24.516 4
RPI 21 -1 1 21 0.522 10.962 19.200 0.478 9.178 20.140 9
SLU 18 0 2 20 0.361 7.220 21.600 0.639 13.802 21.022 7
Union 37 0 0 37 0.554 20.498 25.600 0.446 11.418 31.916 1
Yale 24 0 1 25 0.646 16.150 23.000 0.354 8.142 24.292 5


[Trotsky]

And here's a comparison of Nieu 2015 predicted points with 2014 actuals:

Code Select Expand


[b][b]2014 2015 Diff[/b][/b]
Brown 17 15 -2
Clarkson 24 21 -3
Colgate 29 29        0
Cornell 26 26        0
Dartmouth 16 21       +5
Harvard 16 17       +1
Princeton 8 12       +4
Quinnipiac 28 25 -3
RPI 21 20 -1
SLU 18 21       +3
Union 37 32 -5
Yale 24 24        0


Rounding results in 263 rather than 264 (12*22) points.  The closest team under .5 is Cornell, so if we award Cornell the make up point to make the percentages all work out, we get:

Code Select Expand


[b]2014 2015 Diff[/b]
Brown 17 15 -2
Clarkson 24 21 -3
Colgate 29 29  0
Cornell 26 27 +1
Dartmouth 16 21 +5
Harvard 16 17 +1
Princeton 8 12 +4
Quinnipiac 28 25 -3
RPI 21 20 -1
SLU 18 21 +3
Union 37 32 -5
Yale 24 24  0


[Dutchman]

The seniors on the Union team, Carr, Bodie, Sullivan, Hatch, Ikkala, and the Ghost (jr), refused to accept defeat as an outcome. Their combined leadership cannot be replaced. There is no way we can reload fast enought. We are going to be tough, but doubtful we end up on top of the ECAC. Someone elses turn.

[Trotsky]

The seniors on the Union team, Carr, Bodie, Sullivan, Hatch, Ikkala, and the Ghost (jr), refused to accept defeat as an outcome. Their combined leadership cannot be replaced. There is no way we can reload fast enought. We are going to be tough, but doubtful we end up on top of the ECAC. Someone elses turn.

I'd have hoped so, but the pure numbers say otherwise.  The "problem" is that Union's average over ten isn't bad enough to drop them too far.  Even though they lose a ton, their typical replacement player projects to the second-best in the conference (25.6, second only to Cornell's 27.3). Union's going to regress, but the mean towards which they are regressing is itself competitive.

It could all fall apart, of course (Cornell '13 comes painfully to mind), but if the methodology has any merit Union's natural state just isn't bad enough anymore.  This is unnerving.

The theoretical "0% Returning" rankings:

Code Select Expand

1 Cornell 27.3
 2 Union 25.6
 3 Quinnipiac 23.7
 4 Yale 23.0

 5 Dartmouth 22.4
 6 Colgate 22.0
 7 Harvard 21.9
 8 SLU 21.6

 9 Clarkson 21.4
10 Princeton 19.5
11 RPI 19.2
12 Brown 15.7


[marty]

The seniors on the Union team, Carr, Bodie, Sullivan, Hatch, Ikkala, and the Ghost (jr), refused to accept defeat as an outcome. Their combined leadership cannot be replaced. There is no way we can reload fast enought. We are going to be tough, but doubtful we end up on top of the ECAC. Someone elses turn.

An admirable if not obvious attempt at reverse woofing. May the wrath of those gods have their way.

[Trotsky]

The seniors on the Union team, Carr, Bodie, Sullivan, Hatch, Ikkala, and the Ghost (jr), refused to accept defeat as an outcome. Their combined leadership cannot be replaced. There is no way we can reload fast enought. We are going to be tough, but doubtful we end up on top of the ECAC. Someone elses turn.

An admirable if not obvious attempt at reverse woofing. May the wrath of those gods have their way.

Sandbagging is not a crime.

It is, however, an impiety.